Github-Repo-Org-Question

An ongoing research notebook investigating how to manage a large personal GitHub repo fleet — organisation, navigation, multi-machine sync, bulk lifecycle ops, and launcher UX. Not a one-shot research project: this is a living document that gets picked up whenever a new angle, tool, or pain point surfaces.

Built on Daniel's Claude Research Workspace template — iterative AI-assisted research with Claude Code as the execution engine.

What this notebook is trying to answer

"How do I manage hundreds of personal GitHub repos across multiple machines without wasting time on navigation, losing work to sync accidents, or fighting the filesystem every time folders get reorganised?"

Full pain-point list and learned requirements live in SCOPE.md. Read that first if you're resuming work.

Current state

• Architecture settled (09/04/26): five-layer stack keyed on git remote get-url origin, not filesystem paths. Synthesis in outputs/final/2026-04-09-repo-management-synthesis.md.
• Implementation brief ready for a downstream agent: outputs/final/2026-04-09-implementation-agent-prompt.md.
• Printable PDF of the synthesis: outputs/aggregated/pdf/synthesis.pdf.

Resuming work

Open the repo in Claude Code and run:

/resume

That reads SCOPE.md, the latest outputs/final/, and the most recent individual outputs, then either continues from where the last iteration left off or lets you drop a new question into the queue. To evaluate a specific tool against the spec instead, run /resume evaluate <tool>.

Directory structure

├── CLAUDE.md                    # Instructions for Claude Code
├── SCOPE.md                     # Live spec: pain points + requirements (READ FIRST)
├── context/
│   ├── from-human/              # Background info and notes
│   ├── from-history/            # Compacted findings from prior iterations
│   └── from-internet/           # Saved web sources
├── prompts/run/{initial,subsequent}/   # Prompts, in execution order
├── outputs/
│   ├── individual/              # Per-prompt research outputs
│   ├── aggregated/{markdown,pdf}/
│   └── final/                   # Synthesis + implementation briefs
├── slash-commands/              # Custom Claude Code commands
└── notes/

Slash commands

Command	Purpose
/resume	Pick up from the last iteration, or evaluate a specific tool against SCOPE.md
/run-prompt	Execute the next prompt in the queue
/compact	Summarise individual outputs into compacted history
/aggregate	Combine individual outputs into a single document
/status	Show research progress

Design philosophy

• Filesystem as workflow engine — folder structure defines the process
• Markdown-native — plain text, version-controlled, portable
• Compaction over RAG — summarise and feed back rather than vectorise
• Iterative deepening — each round builds on the last
• Living spec — SCOPE.md is updated every iteration as new pain points and requirements emerge

License

MIT

---

For more Claude Code projects, see the Claude Code Projects Index.